CODEX Digest 10.01.26

Diagnostic stewardship can improve care while potentially reducing unnecessary healthcare spending. This practical guide from the SHEA Diagnostic Stewardship Special Interest Group modeled the financial and quality effects of five stewardship interventions spanning laboratory medicine, infection prevention, blood-culture stewardship, and hospital administration. The authors tested their business-case development approach in a 200-bed hospital, estimating approximately $2.5–$2.8 million would be realized in laboratory and clinical savings.

Demonstrating algorithmic accuracy does not establish that AI improves diagnosis in practice. This analysis shares a five-phase framework for evaluating medical AI, moving beyond fragmented lab-based testing toward real-world clinical benefit. It combines technical, operational, human-AI interaction, clinical, and real-world validation stages, with built-in checkpoints and fallback mechanisms for safety issues or updates. The result is intended to be an adaptable roadmap to help researchers, institutions, and regulators build trustworthy AI systems throughout their lifecycle.

Making diagnostic reasoning explicit can help clinicians understand expert decision-making processes. This commentary describes how clinical problem-solving discussants prepare, frame problems, articulate hypotheses, manage uncertainty, and conduct a “cognitive autopsy.” The authors emphasize that reasoning requires human judgment, context and responsibility—limitations particularly relevant as large language models increasingly generate fluent clinical analyses.

Screening can uncover abnormalities that create diagnostic work without necessarily improving health. This Dutch prospective multicenter study examined incidents and outcomes of incidental findings among 4,122 participants undergoing low-dose CT. Radiologists reported findings in 7.1%, but centralized expert review reduced those requiring primary-care follow-up to 1.8%, and only 0.5% were referred. The results surface opportunities to embed improvement tactics into lung cancer screening programs to limit unnecessary diagnostic cascades.

Diagnostic excellence depends not only on reaching the right diagnosis but also on communicating it effectively. This review examined what “diagnostic disclosure” means in acute care, where patients may otherwise lack understanding of their conditions. The authors identify disclosure as communicating diagnoses in ways that support patient understanding, self-management, and participation in care, thus situating patients as active partners.

Attaining a correct diagnosis is heralded as a team-based process, but periods of focused reflection may also have value. This conceptual perspective, drawing on cognitive psychology, philosophy, and education, explores whether deliberate solitude could support deeper reasoning and reduce cognitive biases to ultimately enhance thinking brought to diagnostic collaboration. The authors emphasize that evidence remains insufficient to recommend a specific practice, making this primarily a hypothesis-generating contribution.

Balancing diagnostic uncertainty with the urgency of sepsis treatment is a persistent patient safety challenge. This observational cohort study of nearly 200,000 emergency room visits at a large urban health system examined relationships between sepsis diagnosis probability and illness acuity. Being aware of the relationship between uncertainty and treatment initiation may help clinicians avoid both premature treatment and potentially harmful delays, although prospective validation is still needed.

Emergency department revisits can reveal missed opportunities for safer diagnosis, but reviewing every return visit is labor-intensive. This pre-print reports on an exploratory single-center study examining whether AI can help broaden quality-review screening. GPT-4 classified nearly all visits as requiring follow-up in contrast to clinicians reviewing the same visit information, raising concerns about over-triggering, while a preliminary knowledge-graph approach showed promise for focusing human review on potentially concerning cases.

Diagnostic technologies can perpetuate inequities when precision varies across patients. In this prospective single-center study of 631 critically ill adults, researchers compared pulse oximeter readings with arterial oxygen saturation using subjective and objective skin-pigmentation measures. The results show that FDA-cleared pulse oximeters tested systematically underestimated arterial oxygen saturation, with less bias in patients with darker skin - in contrast to earlier studies which found that pulse oximetry was less accurate in patients with darker skin. The findings suggest ICU pulse oximeter inaccuracies may overall be more widespread than recognized and continue to warrant attention as a diagnostic equity issue.

Turning diagnostic excellence principles into routine practice remains difficult years after the release of the NASEM Improving Diagnosis in Healthcare report. This rapid qualitative study used seven hospitalist focus groups across 14 hospitals to examine how diagnostic-excellence efforts are developing across largely academic hospital medicine programs. Participants identified organizational culture, resources, leadership, feedback, and practical implementation challenges, offering insight into what may help move diagnostic excellence efforts from the conceptual space into frontline care.

Diagnostic opportunities may emerge long before a formal diagnosis is recorded. This scoping review mapped observational evidence on symptoms, signs, investigations, referrals, and care settings preceding incident heart failure diagnosis in adults. The review highlights how fragmented primary-care signals and diagnostic pathways may contribute to delayed recognition, while cautioning that prediagnostic signals do not necessarily represent missed diagnoses.

Diagnostic delay can be shaped by gaps between screening policy, symptoms, and patient expectations. This qualitative descriptive study interviewed 30 women diagnosed with breast cancer before age 40. The analysis describes patient experiences with symptomatic presentations, delays, emotional and sociocultural challenges, and unmet age-specific needs. The results highlight opportunities for examining patient, clinician, and system contributions to delayed breast cancer diagnosis.